openai / openai/codex

The model reads the same skill multiple times between turns

Open
#39,938 1 comment 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

bug CLI skills
Dominant language
Rust
Stars
125k
Forks
19.4k
PR merge metrics
PR metrics pending

Description

What version of Codex CLI is running?

codex-cli 0.149.0

What subscription do you have?

pro

Which model were you using?

gpt-5.6-sol max

What platform is your computer?

Darwin 25.5.0 arm64 arm

What terminal emulator and version are you using (if applicable)?

ghostty

Codex doctor report

What issue are you seeing?

I have noticed while using codex that codex keeps re reading the same skill over and over between turns.

I think the part that triggers this is in these prompt lines https://github.com/openai/codex/blob/7f9832d0d08c0e2fd56a02421477756474fec95d/codex-rs/ext/skills/src/catalog_prompt.rs#L7-L11

I think specifically Do not carry skills across turns unless re-mentioned

What steps can reproduce the bug?

Here is a simple example

Image Image

You can see from the answer it already knows how to set the model but it goes on and rereads the skill either way because it feels like it needs to read it.

What is the expected behavior?

If the skill is in the context it should not reread the skill it already has it in the context.

Additional information

I tried to make it skip this requirement and it keeps reading the skill because it perceives it as a higher level instruction.

Image Image

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start at codex-rs/ext/skills/src/catalog_prompt.rs, especially lines 7–11, and reproduce the reported multi-turn behavior with the linked example. Trace how “Do not carry skills across turns unless re-mentioned” affects a skill already read in context. Done means an in-context skill is not reread between turns, while explicitly re-mentioned skills still are.

Written by the indexing model from the issue text.

Assessment

Tech stack
rust
Domain
ai-infra-agents, cli
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
55/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.